基于决策逻辑的模糊粗糙神经网络建模  

Modeling by fuzzy rough neural networks based on decision logic

作　　者：叶玉玲 

机构地区：[1]中船重工第七一○研究所仿真与信息研究中心,湖北宜昌443003

出　　处：《计算机集成制造系统》2009年第4期676-680,共5页Computer Integrated Manufacturing Systems

摘　　要：为建立相关量的预测模型,提出了一种新的基于决策逻辑的模糊粗糙神经网络建模方法。首先对原始数据进行预处理,并基于粗糙集理论进行属性约简,得到最简决策表。然后基于决策逻辑建立模糊粗糙神经网络。最后提出了一种结合混沌搜索算法和最小二乘法的Chaos-LS算法,训练模糊粗糙神经网络的参数,从而建立起系统的模糊粗糙神经网络模型。实验证明,这种建模方法建立的模糊粗糙神经网络模型具有较高的精度和泛化能力。A new fuzzy rough neural network modeling method based on decision logic was presented to construct the prediction model of some properties from experiment data. Firstly, the original data was pre-processed and attributes were simplified based on rough set theory to obtain the simplest decision table. Then a fuzzy rough neural network was constructed based on decision logic. Finally, the algorithm Chaos-LS which combined the chaotic search algorithm and least square algorithm was proposed to train the parameters of the fuzzy rough neural network. By this method, a fuzzy rough neural network model was constructed. Simulation results showed that the fuzzy rough neural network model constructed by the modeling method had excellent performance both on accuracy and generalization ability.

关 键 词：模糊粗糙神经网络 决策逻辑 混沌搜索算法 最小二乘法 粗糙集 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]